Back to browse
LocksBet – a price comparison tool for prediction markets

LocksBet – a price comparison tool for prediction markets

by at-w·Jun 10, 2026·2 points·0 comments

AI Analysis

●●SolidSolve My ProblemSlick

Fee-adjusted odds comparison across Polymarket and Kalshi with LLM-verified event matching.

Strengths
  • Hybrid matching pipeline combines embeddings, LLM sanity checks, and human review
  • Fee-adjusted pricing shows real arbitrage opportunities after costs
  • Clean UI with live price gaps and volume data across markets
Weaknesses
  • Prediction market comparison tools becoming crowded with funded competitors
  • Manual review bottleneck doesn't scale as new platforms launch
Category
Target Audience

Prediction market traders, arbitrage seekers

Similar To

OddsJam · Polymarket · Kalshi

Post Description

I built LocksBet to make it easier to see where the same event is trading across prediction markets, and compare the price you would actually get after accounting for fees and liquidity.

The site matches related markets on Polymarket/Kalshi, shows (after-fee) prices side by side, and lets users compare the underlying rules.

Currently, the site uses embeddings and hard gates on stuff like numerical range/date differences to find match candidates. An LLM then sanity checks the embeddings-based match candidates, and sends those that seem like true matches to human review. I then manually review match candidates and use the manually labelled data to improve our automated matching over time.

I'm working on adding more prediction markets as new platforms launch, with the goal of making it easier to find the best available price across venues. I'm also working on showing more complex relationships between events (e.g. one-to-many outcome matches).

It's been fun to work on and I'd love to get any feedback on the site and any other thoughts/questions.

Similar Projects

AI/MLMid

Compare Prices with SupplyFLare AI

Nice, search-focused UI with CSV upload and an explicit LTR angle — that specificity is promising for people who care about ranked pricing signals rather than raw scrape dumps. The site feels early (empty results, minimal onboarding and coverage notes), so the real question is whether their non-stationary-data approach to ranking actually beats simple heuristics at scale. If the ML pipeline and freshness guarantees are solid, this is useful to its niche; right now it's a tidy MVP.

Ship ItSlick
invar1ant
103mo ago